A new AI tool built by researchers in Massachusetts can now show doctors exactly how much radiation travels through a prostate cancer patient's body during treatment — in just 23 seconds. The tool, called DiffuDose, could help tailor therapy to each person instead of giving everyone the same dose.
Right now, most patients receiving a type of treatment called radiopharmaceutical therapy — where cancer-fighting drugs are injected into the body — get a one-size-fits-all amount of radiation. But a team at the University of Massachusetts Amherst says that approach leaves room for improvement.
"Everybody gets the same dose," said Joyita Dutta, a professor in the Riccio College of Engineering at UMass Amherst. "That essentially leaves the therapy's potential untapped. Measuring how much radiation each tissue actually absorbs is the key to personalizing treatment."
Radiopharmaceutical therapy was approved by the FDA to treat late-stage prostate cancer in 2022. Prostate cancer is the second most common cancer in American men, after skin cancer, affecting 1 in 8 men during their lifetimes. For patients whose cancer has spread, the five-year survival rate is around 40 percent.
The challenge is that the best way to measure how radiation spreads through the body — called dosimetry — is accurate but slow. It can take hours per patient, which is too long for doctors to use it in real time.
DiffuDose solves that problem by combining two AI programs. One creates a rough estimate of radiation spread, and the other refines it into a detailed, full-resolution map. The tool matched the gold-standard accuracy in under 23 seconds per patient, and it outperformed six other competing methods across multiple organs including the kidneys and liver.
"Pixel by pixel in a full image, you could see how the dose was distributed across the body," Dutta said. "That's what really helps you personalize the treatment." With this information, doctors could adjust how much drug to give a patient or how often to administer it.
The team published their findings in IEEE Transactions on Radiation and Plasma Medical Sciences. Graduate student Bowen Lei was the first author on the paper. The research also included collaborators from UMass Chan Medical School, Massachusetts General Hospital, and the Institute of Nuclear Medicine in Bethesda, Maryland.
Dutta says her next step is a collaboration with UMass Chan Medical School to combine the AI models with patient blood tests, which could help doctors understand even earlier how well a patient is responding to treatment.